Learn about redlock

What is? RedLock

RedLock yes Redis The official proposal is based on Redis How to implement distributed lock .

Can guarantee

1. Security ： Exclusive access , That is, there will always be only one client Can get the lock
2. Avoid deadlock ： The final client Can get the lock , There will be no deadlock
3. Fault tolerance ： Important parts redis If the node survives, it can provide services normally

Implement distributed locking on a single node

SET key_name my_random_value NX PX 30000  

NX Express if not exist Just set it up and go back True, Otherwise do not set and return False ( Only when the key Set when not present ) PX Indicates the expiration time in milliseconds , 30000 After these milliseconds key Be overdue

Delete own Lock set

Delete reason ： Ensure the high efficiency of server resources , You don't have to wait for the lock to automatically expire before deleting

Delete method ： Best use Lua Script delete （redis Ensure that no other actions are taken when executing this script , Ensure the atomicity of the operation ）, The code is as follows ; Logic is First get key, Delete this if it exists and the value is set by itself key; Otherwise, skip ;

​

if redis.call("get",KEYS[1]) == ARGV[1] then
    return redis.call("del",KEYS[1])
else
    return 0
end

Realization RedLock Algorithm

Algorithm flow , Open five nodes , To claim the lock ,client Operations performed

1. Get the current time in milliseconds
2. client Try using the same key,value Get all redis Service lock , In the process of acquiring a lock, the acquisition time is much shorter than the lock expiration time , Avoid long waits for a closed redis service , And try to get the next instance
3. client By subtracting the time of the first step after obtaining all the locks that can be obtained , This time difference is less than TTL Time and at least three redis Instance successfully acquired lock , Only in this way can we really get the lock successfully
4. If the lock is acquired successfully , Then the effective time of the lock is TTL Subtract the time difference from step 3
5. If the lock acquisition fails , Then it will start to unlock all redis example , Because we have obtained less than 3 A lock , Must be released to prevent impact on other client Get the lock

Failure to retry

When one client When getting lock failed , You should try again later to avoid multiple client At the same time, apply for lock . At best, there is a client At the same time, lock applications are initiated to all services

Lock acquisition failure and lock acquisition failure client Release the lock in time after completing the task

Release the lock

To release a lock, just issue the command to release the lock to all instances , It's not necessary to consider whether the lock can be released successfully ;

Reference article

Redlock（redis Distributed lock ） Principle analysis - RGC - Blog Garden (cnblogs.com)